insertBefore and duplicate tokens

[RunDevelopment]

Right now, the documentation of insertBefore does not specify how the function behaves if the grammar of inside and insert have keys in common.

Example

Let's assume the following language:

Prism.languages.foo = {
    'a': /a/,
    'b': /b/,
    'c': /c/,
    'd': /d/
};

Now, we insert a new pattern for a existing token:

Prism.languages.insertBefore('foo', 'd', {
    'a': /new/
});

// result:
{
    'a': /new/, // old position, new pattern
    'b': /b/,
    'c': /c/,
    'd': /d/
};

And in the other direction: (Prism.languages.foo as originally defined)

Prism.languages.insertBefore('foo', 'b', {
    'd': /new/
});

// result:
{
    'a': /a/,
    'd': /d/, // new position, old pattern
    'b': /b/,
    'c': /c/
};

[RunDevelopment]

I can imagine 3 solutions for this:

1. Throw an error.
   We would detect duplicates and throw if we were to find one. Let the creator of the language decide how this case should be handled.
2. Remove duplicates from inside.
   We remove all duplicates from the grammar of inside and then go on as usual. This ensures that new patterns are both at the correct position and not overwritten by old patterns.
3. Don't insert duplicates.
   If we detect that a token of insert is also present in inside, we ignore it.

I personally prefer option number 2 because it's what I intuitively expected.

Thoughts?